|
Abstract:
|
To measure the similarity between two spatial clusters more reasonably, the special characteristics should be considered during spatial feature selection and feature weighting. One is that the features in one spatial cluster are often not independent of each other, therefore feature structure could provide valuable information. Another one, especially for the purpose of many real applications, is that some goal-related metrics are usually attached to spatial clusters and the similarity measures should align with those pre-defined goals as much as possible. In this paper, two spatial Feature Structure Similarity measures are firstly proposed. Then an optimization approach is employed to incorporate both feature structure and feature value characteristics with proper weights. Finally, the proposed similarity measures are tested in a real banking case for branch investment decision support.
|
